• Welcome to our site! Electro Tech is an online community (with over 170,000 members) who enjoy talking about and building electronic circuits, projects and gadgets. To participate you need to register. Registration is free. Click here to register now.

MAX9611 and the Magic Smoke

Status
Not open for further replies.

vne147

Member
Hello everyone.

I’ve been experimenting with different current monitoring options for a project I’m working. Along the way I discovered the MAX9611/12 which has a built in 12-bit ADC. It seemed like the perfect solution for my needs, so I hooked up a test circuit and wrote some code to interface the IC. Everything seemed to be working fine, that is until I totally fried the part.

I’m still scratching my head about why exactly this happened, so I thought I’d ask ETO to give me a sanity check. I’ve attached a schematic of my test setup.

When the failure occurred, I was supplying the circuit with the Cotek AE-1500-48 switching power supply. I set the voltage and current limits to 48V, and 3.2A respectively. The load was a 2Ω 50W power resistor which was connected to the supply through a digital multimeter to measure current in the 10A range.

I commanded the power supply on and everything was chugging along. The current and voltage measurements were spot on. Then, I unplugged the DMM- lead from the multimeter. I was expecting the measured current to drop to zero and the voltage to jump to 48V. Instead, the IC caught fire and vaporized the RS+ and RS- pins, in the process allowing the mythical magic smoke to escape.

So, what did I do wrong? Can anyone help me solve this mystery?

Any questions or requests for clarification, don’t hesitate to ask. Thanks in advance for any insight you can provide.


MAX9611_Test_Setup.png


Note: The 3.3V supply to the MAX9611 is referenced to GND, not AGND. But I measured the test circuit and there is less than 100 mV difference between the two ground planes. This isn’t how I plan to connect everything in the final circuit. I just did it this way because it was easier to modify my test setup. I don’t think this contributed to the failure, but for the sake of completeness I depicted it in the schematic and since I don’t know what caused the failure, I’m keeping it in my list of possibilities.
 

Attachments

alec_t

Well-Known Member
Most Helpful Member
For a guess, lead inductance caused a switch-off spike greater than the rated 65V at the R pins and the power supply didn't react fast enough to catch that. If that spike caused breakdown within the IC then excess current flowed where it shouldn't and cooked the IC.
 

jamie_s

Member
I reckon when the DMM was disconnected then the voltage present could have been anything upto mains voltage depending on what the rest of the circuit is connected to.
 

vne147

Member
For a guess, lead inductance caused a switch-off spike greater than the rated 65V at the R pins and the power supply didn't react fast enough to catch that. If that spike caused breakdown within the IC then excess current flowed where it shouldn't and cooked the IC.
By lead inductance, do you mean the inductance of the multimeter leads? I know there is no such thing as a purely resistive load, but I wasn't expecting there to be enough inductance in my test circuit to fry this part. Admittedly though, I have very little basis for comparison.

If this was in fact what happened, how do I protect the part from this in the future? I can't guarantee that the end user won't unplug the load just like I did during testing. Do I need some sort of flyback diode?

Thanks.

I reckon when the DMM was disconnected then the voltage present could have been anything upto mains voltage depending on what the rest of the circuit is connected to.
Why would the voltage have been as high a mains voltage? As far as what the rest of the circuit was connected to, there's a microcontroller, some I/O peripherals, and few other bits and pieces that were bypassed via relays at the time, but the 2Ω resistor was the only load connected to the Cotek supply.

Thanks.
 

alec_t

Well-Known Member
Most Helpful Member
The lead inductance is a possibility, not a probability :).
Another possibility is that sudden reduction of the load from 3.2A confused the stabilisation circuitry in the power supply, so that there was overshoot beyond 65V. Do you know what the transient response of the supply is like?
 

vne147

Member
The lead inductance is a possibility, not a probability :).
Another possibility is that sudden reduction of the load from 3.2A confused the stabilisation circuitry in the power supply, so that there was overshoot beyond 65V. Do you know what the transient response of the supply is like?
I haven't done any specific testing on the transient response of the Cotek supply. I only know what's in the datasheet. I remember seeing a test report on it at one time. Maybe that would have something in it about transient response. I'll go see if I can dig it up.

Any ideas on how I can protect the part against this in the future?

Also, after some more poking around I discovered that the H/W I2C module in the MCU was damaged too. Whatever happened with the MAX9611, it was severe enough to bled over into the I2C lines and overvolt the micro pins or damage the hardware in some other fashion. That was an $11 part. Bummer...

EDIT: I was unable to find the test report I mentioned. I may have been thinking about a different power supply I once looking to use.
 

jamie_s

Member
vne147 said:
Why would the voltage have been as high a mains voltage? As far as what the rest of the circuit was connected to, there's a microcontroller, some I/O peripherals, and few other bits and pieces that were bypassed via relays at the time, but the 2Ω resistor was the only load connected to the Cotek supply.

Thanks.
because of ground loops and the cotek having no reference to AGND. you dont know how high the voltage could be floating on the +48v in respect to AGND and your MCU power suppply
 

vne147

Member
because of ground loops and the cotek having no reference to AGND. you dont know how high the voltage could be floating on the +48v in respect to AGND and your MCU power suppply
I think I understand why you were asking what else was connected to the circuit now. If there was a lot of current flowing through the .1Ω resistor separating the GND and AGND planes, then there would have been a significant voltage drop across it which would have added to the Cotek supply's output. Is my thinking correct here?

As far as what else was referenced to AGND and drawing current, there were a bunch of analog and mixed signal ICs. I'd estimate the total current flowing from AGND to GND to be less than 50 mA. That should have only caused a potential difference between GND and AGND in the low millivolt range, which I measured to be the case. But I could be wrong on my current estimate, or the circuit may have a fault I haven't detected yet.

With respect to the ground loop thought, AGND is connected to GND at only one point. The 2 caps and the resistor are placed immediately adjacent on another. I need to do a little bit more thinking about that, but do you think this is a real possibility?

Anyway, in theory if the factor of a floating 48V input with respect to AGND contributed to the failure, that could be mitigated by just connecting the MAX9611 to GND instead of AGND. Do you agree?

Thanks.
 

alec_t

Well-Known Member
Most Helpful Member
A TVS on your equipment would go some way to protecting it.
 

vne147

Member
A TVS on your equipment would go some way to protecting it.
I've never used a TVS before. My local electronics shop had these in stock though: NTE4951. They were the closest in stock that fit my needs. What do you think?

Also, I'm not 100% sure how they should be connected. I took a stab at it and attached an updated schematic. Am I using them correctly?

Thanks.

MAX9611_Test_Setup_1.1.png
 

alec_t

Well-Known Member
Most Helpful Member
Connection looks fine. The 4951 has a rated clamping voltage of 70.1 (i.e above the magic 65V), but a breakdown voltage of 51.05. I'm not sure how to interpret those figures to say if that would be suitable in your application. Someone with experience of TVS use will hopefully chime in. What is the maximum supply voltage your circuit should see?
 

vne147

Member
Connection looks fine. The 4951 has a rated clamping voltage of 70.1 (i.e above the magic 65V), but a breakdown voltage of 51.05. I'm not sure how to interpret those figures to say if that would be suitable in your application. Someone with experience of TVS use will hopefully chime in. What is the maximum supply voltage your circuit should see?
The maximum voltage I'll ever intentionally command the power supply to output will be 48V. That's what the supply was commanded to when the fault occurred. It was also current limited to 3.2A at the time.

So, with the 2Ω load I had, before the fault I was drawing about 3.2A @ 6.4V. When I disconnected the multimeter lead, I was expecting the voltage to jump to 48V but instead I got smoke. The rest, as they say, is history.

Maybe the supply has a poor transient response as you suggested could be the cause. Or maybe it just overshot too much, which in my mind would fall under the poor transient response umbrella. According to the Cotek's datasheet, the over voltage protection is supposed to kick in at 120% of 48V or 57.6V. That's why I thought I'd be OK with a 65V part.

I have no idea how high the voltage got because it all happened so quickly. But, I'd venture a guess that it was way above 65V otherwise we wouldn't be having this conversation.

Any other thoughts or considerations I should be making?

Thanks for your help.
 

spec

Well-Known Member
Most Helpful Member
Hi vne,

Your problem sounds very odd because for one thing those current monitoring chips are pretty robust- they have to be. But the other thing is that the current monitoring input side is buffed by some high value resistors to get the high input voltage capability.

I know this sounds like a cop-out, but I would suspect some fundamental cause:
(1) The chip was faulty
(2) Something got shorted while you were disconnecting
(3) Your circuit is in fact not what you think it is.

In my vast experience of blowing chips up you get some odd relationships, like the circuit works fine, except on Tuesday morning (dry joint). When we moved to a new lab a board in the equipment blew (wrong value resistor fitted). Whenever the customer came for a demonstration the on- board processor would lock up (pin on EPROM not connected) and my favorite: every time you turned left in an automobile, the front drivers side wheel screeched (radiator electric fan bearing dry).

So there it is for what it is worth.

The way forward would seem to be to measure what your power supply does when you connect and disconnect the resistor.

Build another circuit with a new chip and test gently, gently. Say start with 10V and a few hundred mili amps and gradually build up.

Finally, there is the catchall fault: your circuit may be oscillating. That would give exactly the symptoms you describe. The solution is to have a good layout with plenty of ceramic decoupling capacitors in the correct positions. Those current sensing amps can be a touch twitchy in the frequency domain.:)

I will get my coat now!

spec

PS: perhaps post an image of your circuit for us to have a look at!
 
Last edited:

alec_t

Well-Known Member
Most Helpful Member
The way forward would seem to be to measure what your power supply does when you connect and disconnect the resistor.
Good suggestion. If you do that, set up the same voltage and current as before, remove the IC to stop it being fried, then disconnect your meter lead, you could monitor/record the supply transient response with a 'scope.
 

spec

Well-Known Member
Most Helpful Member
Good suggestion. If you do that, set up the same voltage and current as before, remove the IC to stop it being fried, then disconnect your meter lead, you could monitor/record the supply transient response with a 'scope.
Then the roof of the building would cave in or you would be struck by lightning. I had the former happen to me.:arghh:

Faults are God's way of showing who is in charge. It always amazes me how often you get two totally unrelated faults at the same time. You change an auto wheel and then the lights fail and there is no relationship.

spec
 

alec_t

Well-Known Member
Most Helpful Member
You change an auto wheel and then the lights fail and there is no relationship.
Depends if you believe in Chaos Theory or not. If a butterfly flapping its wings in Africa can cause a hurricane in the Bahamas, then surely the lights failing in response to changing the wheel would not be surprising? :).
 

spec

Well-Known Member
Most Helpful Member
Depends if you believe in Chaos Theory or not. If a butterfly flapping its wings in Africa can cause a hurricane in the Bahamas, then surely the lights failing in response to changing the wheel would not be surprising? :).
:D Too true Alec. It always amazes me how nature works- someone somewhere must have a sense of humor!

Also, it is surprising how fallible humans (including me) are, in diagnosing faults.

The number of times that people have spent days investigating a fault when, all along, it has been a simple PSU problem. In the old days with TVs the diagnosis was inevitably, the tube, followed by the picture valve (vacuum tube) when, in my experience, electrolytic capacitors and power resistors were the main culprits (paragraph not for Nigel consumption).:)

I feel more stories cumming on, so I will get my coat.:)

spec
 
Last edited:

vne147

Member
(1) The chip was faulty
Possibly. Although I only had 2 to start with, more are on the way. For the first one to be faulty, that seems quite unlucky from a statistical standpoint. It certainly would be consistent with other experiences in my life though.

(2) Something got shorted while you were disconnecting
I'm not going to say this is an impossibility, but I'm pretty sure this didn't happen.

(3) Your circuit is in fact not what you think it is.
This is a distinct possibility. But, if this is the case I can't figure out for the life of me where my perception and reality diverge.

The way forward would seem to be to measure what your power supply does when you connect and disconnect the resistor.
Good suggestion. If you do that, set up the same voltage and current as before, remove the IC to stop it being fried, then disconnect your meter lead, you could monitor/record the supply transient response with a 'scope.
This is a good suggestion. Unfortunately, I don't have suitable equipment at my disposal to perform these tests.

perhaps post an image of your circuit for us to have a look at!
The circuit and board are quite complicated, and I would rather not post it to a public forum such as this. However, if you and/or Alec are willing to invest a little time looking as it, I'll gladly PM it to you. Please be gentle though. I am not an EE.

If we all arrive at a solution, I'll post the details back here for posterity. But I'd prefer to not have the complete design floating around the interwebs. If it made it to Facebook, some people might confuse it for news and elect it president.

Then the roof of the building would cave in or you would be struck by lightning.
I thought I heard the house settling immediately prior to this chip toasting. Maybe it was just my imagination.

P.S. I'm sorry you had a roof fall on of your head.

Faults are God's way of showing who is in charge. It always amazes me how often you get two totally unrelated faults at the same time. You change an auto wheel and then the lights fail and there is no relationship.
I require precious few reminders to show me I'm not in charge of anything, let alone what an electron decides it wants to do. In my head, I contemplate nearly all principles of electric theory by imagining hand drawn diagrams featuring a cartoon electron named Sparky. Translation: I know I know nothing.

Depends if you believe in Chaos Theory or not. If a butterfly flapping its wings in Africa can cause a hurricane in the Bahamas, then surely the lights failing in response to changing the wheel would not be surprising? :).
Just thinking about this caused a disturbance in the ambient EM field around my head which I'm slightly certain caused a fly to buzz it's way into my ceiling fan just now. I submit that the word "unrelated" should be stricken from the English language as there is no cause to ever use it whilst describing actual relationships or events.


Thank you gentlemen for your help.
 

spec

Well-Known Member
Most Helpful Member
Possibly. Although I only had 2 to start with, more are on the way. For the first one to be faulty, that seems quite unlucky from a statistical standpoint. It certainly would be consistent with other experiences in my life though.
Stranger things have happened at sea... and in the lab.

Unfortunately, I don't have suitable equipment at my disposal to perform these tests
Try connecting a one Ohm resistor in series with the PSU output and then connect a large electrolytic capacitor from the free end of the resistor to 0V. Then use the capacitor as your PSU.

However, if you and/or Alec are willing to invest a little time looking as it, I'll gladly PM it to you.
That would be fine by me

If it made it to Facebook, some people might confuse it for news and elect it president.
:)

I'm sorry you had a roof fall on of your head.
A whole roof section blew off in a gale. I was picking shards of glass and tile out of equipment and documents for months after that.:arghh:

I require precious few reminders to show me I'm not in charge of anything, let alone what an electron decides it wants to do.
Electronics is not natural. It is the gods trying to tell you something.

In my head, I contemplate nearly all principles of electric theory by imagining hand drawn diagrams featuring a cartoon electron named Sparky. Translation: I know I know nothing.
Yes, visualizing things is useful. I always liken electrons to water.

spec
 
Status
Not open for further replies.

Latest threads

EE World Online Articles

Loading
Top